Research On Recognition Of Orchard Diseases And Pests Based On Deep Learning

Agricultural diseases and pests do great harm to crops and their products.Many diseases and pests have wide distribution range,large quantity and fast reproduction speed.Only by monitoring the situation of diseases and pests as soon as possible can we make prevention and control in time.Therefore,aiming at the most common problems in the process of orchard pest object detection and classification,the deep learning method of intelligent recognition of common typical pests and diseases is researched,and the effectiveness of the proposed model and method is verified by experiments.The main research contents of this paper can be summarized as follows:1.Aiming at the problem that the accuracy of the existing object recognition method needs to be improved,based on the establishment of the orchard pest data set in this paper,by improving the structure of convolution neural network and feature extractor,and using the technology of data enhancement and data reorganization,a new method of pest recognition and counting based on MPest-RCNN improved by Faster R-CNN is proposed.The method solves the problem that the recognition rate of multiple pests is low due to the large individual difference when they are recognized at the same time.The experimental results show the effectiveness of the orchard pest recognition and counting method based on MPest-RCNN deep learning model,and the recognition accuracy of this method reaches 99.11%.2.In the classification process of convolutional neural network,there is a certain degree of over fitting phenomenon and the poor ability of traditional classifier to deal with nonlinear problems.By improving activation function,adding residual module and dropout layer,the over fitting problem of convolutional neural network classification model in the training process is effectively alleviated.By improving the classifier,a better nonlinear classification effect is achieved.A new method of ERes Net-SVM for orchard pest classification based on Res Net18 improved residual network is proposed,which solves the problems of over fitting and convolution neural network feature extraction in the training process.The experimental results show that the recognition accuracy of ERes Net-SVM for 8 diseases is 99.3%,which is 5.9% higher than the original model,and the recognition accuracy of 6 pests is 100%,which is 3.9% higher than the original model.3.Aiming at the problem of low recognition efficiency of traditional pest recognition system,based on the establishment of typical orchard pest recognition and counting model,an orchard pest recognition and counting system based on deep learning is developed and designed through Python language and Django framework.The system can operate and run easily and efficiently on the computer,which provide theoretical and technical reference for the promotion and application of intelligent agriculture and precision agriculture in the future.